A new mechanical method for creating reactive laminate powders has been developed using a two-step process; in the first step bulk reactive materials are created by cold-rolling stacks of alternating sheets of nickel and aluminum into foils with bilayer thicknesses ranging from 2.9 to 1.8 mu m. This step establishes the average reactant spacing and, hence, the reactivity of the material. In the second step the rolled foils are then ground into laminate powders and sieved based on their diameters, which range from 850 to 53 mu m. Our processing methodology allows the particle size and the reactant spacing to be varied independently. Powders made by this method have heat releases within a differential scanning calorimeter (DSC) that vary with the average reactant spacing, similar to rolled and sputter deposited foils. However, the measured heats also vary with the average diameter of the powders, as smaller particles show a systematic decrease in heat. Furthermore, this effect is magnified for the powders with the coarsest microstructure, as they show the largest drop in DSC heat release. The physical densities also vary as a function of particle size. The powders with the largest average bilayer thickness become Ni-rich at the smallest particle sizes, powders with the next finest average bilayer thickness become Al-rich, and powders with the smallest average bilayer thickness show little variation. We attribute the particle size dependence of the DSC heats to small powders being broken from regions of the original rolled foils that contain a high volume fraction of Ni-rich and Al-rich bilayers. These microstructural and chemical variations alter the exothermic reactions that are seen during slow heating in a DSC, as well as the heats of reaction that are measured in the DSC as a function of powder size. We support this hypothesis of non-random breakup during grinding by simulating bimodal distributions of bilayer chemistry within powders and modeling their densities and heats of reaction. The simulations are compared with measured values. Finally, we normalize the effects of particle size and bilayer thickness by plotting the measured DSC heats of reaction versus the number of bilayers per particle; the values merge toward one curve, with the largest decrease in heat occurring when there are fewer than 150 bilayers in each particle. This ratio proves useful when selecting particles for particular applications.